Paraspeech for Mac developers
Speak the context. Check the constraints.
Draft your next AI coding prompt with your voice. Say the bug, scope and acceptance criteria, then check the text before sending it to your coding tool.
Download for MacmacOS 14.0 or later. Local transcription requires Apple Silicon and model setup. Intel Macs require cloud models and a subscription.
Installing later? Open paraspeech.com/for/developers on your Mac.
From spoken request to a prompt you can send
Step 1
Say the intent
Start dictation and explain the problem, what can change, what must stay the same, and how to check the result.
Step 2
Check the details
Review the transcript. Type exact file paths, identifiers and commands, and check that every constraint survived.
Step 3
Send when ready
Use the reviewed text in your coding tool. If its input field does not accept automatic insertion, paste from the clipboard.
Try a prompt with real constraints
We tested a synthetic bug-fix request with fictional files. The raw transcript reached the clipboard, but it needed edits before it matched the request. In particular, “unchanged” was transcribed as “on change”.
Raw transcript
Writeabuckfix for duplicate retry callbacks in SRC slash network slash retry dot ts. Keep the public API on change. Keep the ext retry limit on change. Add a regression test in test / network / retry. The test must fail before the fix and test afterwards. Explain the cause of the duplicate callback. Make the smallest change that fixes it, and report which test you ran.
After manual review
Write a bug fix for duplicate retry callbacks in src/network/retry.ts. Keep the public API unchanged. Keep the existing retry limit unchanged. Add a regression test in tests/network/retry.test.ts. The test must fail before the fix and pass afterward. Explain the cause of the duplicate callback, make the smallest change that fixes it, and report which test you ran.
- Corrected the opening to "Write a bug fix".
- Typed the exact source path: src/network/retry.ts.
- Changed "public API on change" to "public API unchanged".
- Restored "existing retry limit unchanged".
- Typed the complete test path: tests/network/retry.test.ts.
- Changed "test afterwards" to "pass afterward".
Check both file paths, the unchanged public API, the unchanged retry limit, and a regression test that fails before the fix and passes afterward. Text appearing in a field is only the start; the prompt needs to say what you meant.
How we ran this example
September 29, 2026: a Debug build of Paraspeech 1.7.3 Mac source (build 372) with the workspace's native packages, not the signed public app. Apple M4 Pro, macOS 27.0, Local Multilingual S transcription and rewriting off. Prerecorded synthetic speech played through the speakers into the built-in microphone. Models and microphone permission were already ready, and test access was enabled.
This is one microphone-to-clipboard example, not a real-voice accuracy or speed benchmark. It does not test first-time setup, free usage, checkout, or insertion into a coding app. No prompt was sent to an AI service.
Read the spoken input, output and test detailsChoose where transcription happens
Local transcription runs on Apple Silicon after model setup. If you allow cloud processing, it can be used when local processing is unavailable. Cloud operations send the required audio or text to their providers.
Your coding tool has its own data practices. Transcribing locally does not make an AI service local when you send it the prompt.
Read about processing choicesTry the workflow, then choose a plan
Download the app and set up dictation. Free transcriptions before plan selection are separate from the seven-day subscription trial offered at checkout.
Subscriptions include local and cloud models. Lifetime covers local models, includes twelve months of updates, and lets you keep using the last covered version. Further updates may require a new purchase.
See current prices and purchase termsUse the same approach around your code
The problem, change and verification also make a useful starting point for an issue description or PR summary. For more examples, read our developer dictation workflow guide. If you are comparing products, see our Paraspeech and Wispr Flow comparison, prepared by the Paraspeech team.